QRP01-4: Analysis of Single Hop Ad-hoc Networks

The information theoretic principles of maximum entropy (ME) and minimum relative entropy (MRE) are employed, as least-based methods of inference, for the analysis of single ad hoc systems with bursty arrivals and blocking. The system consists of MAC and IP layer finite capacity links together with a shared medium which forwards from the MAC to the higher IP layer links in a determined fashion. The traffic at each MAC layer link is assumed to be bursty and is modeled by a compound Poisson process with geometrically distributed batches. The shared medium provides generalized exponential (GE) type of service whilst each output link forms a single GE-type FCFS server queue. To this end, an approximate MRE decomposition algorithm is proposed for the partitioning of the shared medium with its associated MAC layer links into ordinary finite capacity virtual queues with server vacations and modified service times. Consequently, each queue is analyzed in isolation by exploiting a stochastic relationship between polling time and server vacation. Moreover, the IP layer links are analyzed separately as GE-type finite capacity queues. Finally, the individual building block queues, are combined together via an iterative process which is based on a MRE solution for arbitrary open queuing network models with blocking and GE-type flow formulae.